Determining the 3D genome structure of a single mammalian cell with Dip-C

Summary 3D genome structure is highly heterogeneous among single cells and contributes to cellular functions. Our single-cell chromatin conformation capture (3C/Hi-C) technique, Dip-C, enables high-resolution (20 kb or ∼100 nm) 3D genome structure determination from single human and mouse cells. Dip-C is robust, fast, cheap, and does not require specialized equipment. This protocol describes using human and mouse brain samples to perform Dip-C, which has also been applied to other tissue types including the human blood and mouse eye, nose, and embryo. For complete details on the use and execution of this protocol, please refer to Tan et al. (2021).


MATERIALS AND EQUIPMENT
Reagents to prepare for isolation of nuclei from the brain 1.5 M sucrose. Dissolve 20.538 g sucrose (Sigma 84097) in water and adjust volume to 40 mL. Filter and store indefinitely at 4 C. For long-term storage, aliquot and store indefinitely at À20 C to avoid bacterial growth.
Note: If used regularly, 1.5 M sucrose may develop white bacterial growth after a few months at 4 C. If so, discard and use another unopened aliquot.
Nuclei Isolation Medium 1. Each reaction consumes 6 mL. The following recipe (45 mL) is sufficient for 7 reactions: Optional: If using cells rather than nuclei, reagents to prepare for chromatin conformation capture: a 10% Igepal CA 630. Each reaction consumes 20 mL. The following recipe (1 mL) is sufficient for 45 reactions: Reagents to prepare for whole-genome amplification 60 mg/mL Qiagen Protease. Dissolve 1 vial (7.5 AU) of Qiagen Protease (Qiagen 19155) in 2.78 mL water. Filter, aliquot, and store indefinitely at 4 C. 14.3 mM (5 mg/mL) DAPI. Dissolve 1 tube (10 mg) of DAPI (ThermoFisher D1306) in 2 mL water. Aliquot and store indefinitely at 4 C. 0.1 X TE. The following recipe is for 40 mL: 100 mM Carrier ssDNA. Dissolve each 1 nmol Carrier ssDNA (Table 1; IDT; standard desalting) in 10 mL 0.1 X TE to a final concentration of 100 mM. Store indefinitely at À20 C. 12.5 mM Nextera i5 and i7 Primers. Dissolve each 1 nmol Nextera i5 or i7 Primer (Table 1; IDT; standard desalting) in 80 mL 0.1 X TE to a final concentration of 12.5 mM. Store indefinitely at À20 C.

STEP-BY-STEP METHOD DETAILS
Isolation and fixation of nuclei from the human or mouse brain

Timing: 2 h
In this step, cell nuclei are isolated from the human or mouse brain through mechanical homogenization (Dounce homogenizer) in the presence of a detergent (0.1% Triton X-100), and preserved with a fixative (2% PFA).
Isolation of nuclei was adapted from Lacar et al., 2016) with minor modifications. In particular, Tris buffer was replaced with an equal molarity of HEPES buffer to avoid interference with PFA fixation.
Note that nuclei isolation may alter native 3D genome structure, because the cytoskeleton and gene transcription may be disrupted. Nuclei should be isolated as fast as possible and kept at 4 C until fixation to minimize changes to the 3D genome. We have not tested fixation before nuclei isolation, because homogenization is generally more challenging for fixed tissues.

Nuclei isolation
Optional: If using cells rather than nuclei, skip this section and proceed directly to Fixation (Step 14). Note: The nuclei suspension may be cloudy because of debris (e.g., myelin). Debris does not affect downstream procedures, and will be partially solubilized during SDS treatment (Step 26 and Step 27) at the Chromatin Conformation Capture step.

Fixation
Optional: If using cells rather than nuclei, start from here.
14. Freshly prepare 1% BSA in PBS: Dissolve 0.1 g BSA (Gemini 700-106P) in 10 mL PBS (Thermo-Fisher 10010023). Each reaction consumes 1.2 mL. Chill on ice. CRITICAL: PFA is hazardous. Perform the above and following steps (until Step 19: resuspension of the pellet in 1% BSA in PBS) in a fume hood and properly dispose of waste.
Note: We have not tested other types of formaldehyde (e.g., methanol-containing), other fixatives, or unfixed cells or nuclei.
Note: BSA, rather than the more widely used glycine, is used to react with excess PFA because in our hands, reaction between glycine and PFA acidifies the solution (yellow when phenol red is present, indicating pH < 6) and would dissolve all cells or nuclei if left for too long (> 1 hour on ice).

Note:
The above step is not aimed to fully quench PFA. Addition of BSA greatly reduces loss of cells or nuclei by preventing cells or nuclei from sticking to the side of the tube, and from aggregating when spun down and resuspended in 1% BSA in PBS.
18. Centrifuge at 1000 g for 5 min at 4 C.
Optional: If using cells rather than nuclei, centrifuge at 600 g instead.
Note: The above step fully quenches PFA, and was adapted from Thomsen et al. (2016). 20. Measure cell or nuclei density with a disposable hemocytometer (INCYTO DHC-N01; manufacturer's protocol: http://www.incyto.com/shop/item.php?it_id=1482380591) and optionally Trypan Blue (ThermoFisher 15250061) if debris is abundant. 21. Aliquot up to 500 k-1 m cells or nuclei per tube. There is no lower bound in principle; see before you begin for details. Each adult mouse brain approximately corresponds to 8 tubes for the cortex and 2 tubes for the hippocampus (2 sides combined). Too many cells (> a few million) may lead to insufficient digestion/ligation and aggregation of cells or nuclei. 22. Centrifuge at 1000 g for 5 min at 4 C.
Optional: If using cells rather than nuclei, centrifuge at 600 g instead.
Note: The pellet may be large because of debris. Debris does not affect downstream procedures.
Pause point: Fixed cells or nuclei can be stored indefinitely at À80 C.

Chromatin conformation capture (3C/Hi-C)
Timing: 2 days (or shorter, if using commercially available kits) In this step, after detergent (0.5% SDS) treatment, chromatin in fixed nuclei is digested with restriction enzyme(s) (e.g., MboI, DpnII, and/or NlaIII), and re-ligated with a DNA ligase to form artificial linkages (i.e., ''chromatin contacts'') between genomic loci that are far away along the linear ll OPEN ACCESS STAR Protocols 2, 100622, September 17, 2021 sequence but nearby in the 3D space. Success of digestion and ligation is assessed by extracting DNA from a small portion (5%) of the reaction, and measuring its length distribution.
This step was adapted from (Nagano et al., 2017;Rao et al., 2014), and can be replaced with other 3C/Hi-C protocol or commercially available 3C/Hi-C kits such as the Arima-SC kit.
Digestion 24. Thaw 500 k-1 m fixed cells or nuclei on ice.
Optional: If starting from fixed cells rather than nuclei, perform the following additional steps: a. Prepare Hi-C Lysis Buffer. Each reaction consumes 1 mL: b. Freshly prepare Hi-C Lysis Buffer with Inhibitor. Each reaction consumes 600 mL: c. Resuspend cells in 600 mL ice-cold Hi-C Lysis Buffer with Inhibitor. d. Incubate on ice for 15 min, occasionally inverting the tube. e. Centrifuge at 2500 g for 5 min at 4 C. f. Remove supernatant. Resuspend in 500 mL ice-cold Hi-C Lysis Buffer. g. Centrifuge at 2500 g for 5 min at 4 C.
Note: SDS treatment is necessary to obtain a large number of contacts per cell. Without SDS treatment, the number of contacts per cell may decrease by 2 orders of magnitudes. 30. Add restriction enzyme(s) and buffer: 25 mL 10 X NEBuffer 2 (NEB B7002S) and 20 mL 25 U/mL MboI (NEB R0147M). Alternatives include: 25 mL 10 X CutSmart Buffer and 20 mL 10 U/mL NlaIII (NEB R0125L), 25 mL 10 X NEBuffer DpnII and 10 mL 50 U/mL DpnII (NEB R0543M), or a combination of multiple enzymes. 31. Rotate at 37 C for 1-24 h. 32. Take 5% (13 mL out of the total 265 mL) and store at 4 C as a Digestion Control.

Reagent Final concentration Amount
Ligation 33. Centrifuge at 1000 g for 5 min at 4 C. 34. Freshly prepare Ligation Buffer. Each reaction consumes 2 tubes. The following recipe (1 tube) is sufficient for 0.  Figure 1 for representative Bioanalyzer traces.

Whole-genome amplification (WGA) by tagmentation
Timing: 1 day In this step, single cell or nuclei are sorted into multi-well plates, lysed, and amplified with transposition (Tn5) and PCR.
The procedure below describes amplification with our implementation of the Illumina Nextera chemistry. If higher sensitivity is required-for example, when the number of contacts obtained is insufficient for distinguishing desired cell types or for 3D modeling with desired spatial resolution, please follow the procedure of our multiplex end-tagging amplification (META) method (Tan et al., 2018). META can detect 2 times as many contacts as Nextera, but involves custom Tn5 transposomes and 2 additional PCR steps.
For first-time users, we recommend starting with 1 96-well plate. Experienced users may amplify 4-8 plates at a time, depending on the number of available PCR machines.
Nextera Index Primers listed in the key resources table allow the pooling of up to 384 cells or nuclei to be sequenced on the same lane. Other index designs may allow more cells or nuclei to be pooled (e.g., 10-bp dual Nextera indices from IDT allows 3,840). Note: Addition of Carrier ssDNA reduces loss of input DNA materials by preventing genomic DNA from sticking to the side of the tube, especially in PCR tubes that are not low-retention.
Note: Volume (0.25 mL) of 60 mg/mL Qiagen Protease does not need to be exact. If desired, however, pipetting accuracy can be increased by freshly diluting 60 mg/mL Qiagen Protease prior to addition (e.g., 1:100 dilution followed by the addition of 25 mL instead of 0.25 mL).
Optional: Before the addition of 60 mg/mL Qiagen Protease, Dip-C Lysis Buffer can be stored indefinitely at À20 C.
53. Add 2 mL Dip-C Lysis Buffer per well to a DNA low-bind 96-well plate (semi-skirted: Eppendorf 0030129504; or skirted: Eppendorf 0030129512, depending on the FACS and PCR machines).
Note: To maximize speed, use a 12-channel pipette to add solution to each well in the above and all subsequent steps.
Pause point: Sorted cells or nuclei in Dip-C Lysis Buffer can be stored on ice for a few hours before lysis.
57. Lyse the cells by running the following PCR program: Figure 2. Representative flow cytometry diagrams with 2 roughly equivalent gating strategies The minor fraction of particles with double, triple, or even higher DAPI signals (''V450-A'') were aggregates from the Chromatin Conformation Capture step. Both were run on a BD FACSAria flow sorter. The 2 gating strategies arose from personal preferences of different flow cytometer operators, and do not affect the results. Note that we primarily study cells in the G0/G1 phase of the cell cycle; the corresponding gate (e.g., ''G1'' in (B)) should be adjusted when studying other phases of the cell cycle.

Hold 4 C Forever
Store at À80 C.

OPEN ACCESS
Pause point: Lysed cells or nuclei can be stored for a few months at À80 C.
Note: The film (Bio-Rad MSB1001) may peel over time at À80 C. This can be avoided by changing to cold-resistant film (Bio-Rad MSF1001) after lysis.
Optional: For longer-term storage, cells or nuclei can be sorted into empty 96-well plates rather than Dip-C Lysis Buffer, and stored indefinitely at À80 C.
Transposition 58. Prepare Transposition Buffer. Each well consumes 8 mL. The following recipe (1 mL) is sufficient for 1 96-well plate: Optional: Transposition Buffer can be stored indefinitely at À20 C.
59. Freshly prepare Transposition Mix. Each well consumes 8 mL. The following recipe is sufficient for 1 96-well plate (with 10% overhead): CRITICAL: The amount of Tn5 transposome per well ($0.015 mL above) determines the length of the final sequencing library. It should be titrated in a pilot experiment with a concentration gradient of Tn5 transposome to obtain an average length of $500 bp. Please refer to troubleshooting 2 for details, and Figure 3 for representative Bioanalyzer traces.
Optional: For first-time users, a Positive Control well can be set up as 2 mL of a 5 pg/mL dilution of any genomic DNA (e.g., diluting 100 ng/mL HeLa gDNA (NEB N4006S) 1:20,000 in water). A Negative Control well can be set up as 2 mL water. Please refer to troubleshooting 1 for details.
60. Add 8 mL Transposition Mix per well (total volume: 10 mL), avoiding touching the liquid (i.e., pipette onto the side, rather than the bottom, of the well). Vortex and spin down. Note: Before PCR amplification, we typically avoid touching the liquid with pipette tips to minimize loss of input DNA materials. In particular, if pipette tips touch the liquid, genomic DNA may stick to the tips and get lost when tips are withdrawn from the liquid. However, the efficacy of this precaution has not been tested systematically; touching the liquid may be acceptable if the resulting data is satisfactory.

Reagent Final concentration Amount
61. Transpose the genome by running the following PCR program: Range suitable for sequencing is indicated by a dashed green box ("acceptable''), and the optimal concentration shown by a solid green box (''best''). All were run on a Bioanalyzer High Sensitivity DNA kit.
Dip-C transposition. Lid temperature: 60 C. Volume: 10 mL Step Temperature Time Cycles (i.e., pipette onto the side, rather than the bottom, of the well). Arrange the indices so no cells share the same index on each sequencing run; see Figure 4 for an example arrangement.
67. Add 11 mL PCR Mix per well (total volume: 25 mL per well), avoiding touching the liquid (i.e., pipette onto the side, rather than the bottom, of the well). Vortex and spin down.
68. Amplify the genome by running the following PCR program: Note: The above PCR program consists of 14 cycles, which is suitable for the input DNA amount of human and mouse samples ($6 pg per cell or nucleus, given a diploid genome size of $6 Gb). The number of cycles may need adjustment if an organism has a very different genome size.
Pause point: PCR reactions can be stored on ice for a few hours, or indefinitely at À20 C.

Purification and size selection
69. Pool all wells from a 96-well plate.  CRITICAL: Avoid cross contamination of liquid between plates that use overlapping indices.
Note: Each 96-well plate (total volume: 2.4 mL) can be pooled directly into 12 mL DNA Binding Buffer and vortexed (total volume: 14.4 mL). Because each PCR purification column can only load 800 mL at a time, we typically use 6 columns per plate to save time; each column only needs to be loaded 3 times. After loading and washing, elute each column into 66.7 mL TE (ThermoFisher AM9849) and pool (total volume: 400 mL).
Optional: For the Positive Control and Negative Control, each well (25 mL) is mixed with 125 mL DNA Binding Buffer. Elute each into 6 mL TE (ThermoFisher AM9849).
Pause point: Purified libraries can be stored indefinitely at À20 C.
Note: The remaining half ($200 mL) serves as a back-up in case size selection or sequencing fails. Depending on sample submission requirements (i.e., minimum DNA amount) of the sequencing provider, the above purification and size selection steps can be scaled down (e.g., to half of the volume), with any extra liquid stored indefinitely at À20 C in the form of a mixture of PCR reactions and DNA Binding Buffer.
Pause point: Final libraries can be stored indefinitely at À20 C. 73. Measure DNA concentration with a Qubit 13 dsDNA HS Assay. Measure DNA lengths with a Bioanalyzer High Sensitivity DNA kit (or Fragment Analyzer). To evaluate the results, please refer to Expected Outcomes for details and Figure 5 for representative Bioanalyzer traces.
Note: To saturate the sequencing library, we sequence each cell with 3-6 m read pairs.
Note: If 3D reconstruction of diploid genome structures (by reading heterozygous SNPs) is not required, shorter read lengths (e.g., paired-end 75 bp) can be used.
Optional: Before deep sequencing, the presence and prevalence of chromatin contacts can be tested at almost no cost with as few as 1,000 reads per plate (e.g., on a MiSeq), which allows the calculation of the ''contact rate'' CR and the ''contact density'' CD; see quantification and statistical analysis for details.

EXPECTED OUTCOMES
For isolation and fixation of cell nuclei, we typically obtain 6 m fixed nuclei from the mouse cortex, 1.5 m from the mouse hippocampus (2 sides combined), 40 m from the mouse cerebellum, 2 m from 100 mg human cortex, and 40 m from 100 mg human cerebellum.
For chromatin conformation capture (3C/Hi-C), the Digestion Control should have a typical length of 2 kb on a Bioanalyzer High Sensitivity DNA chip, whereas the Ligation Control should have a peak around the upper marker (10 kb) (Figure 1). If starting from 500 k cells or nuclei (i.e., 500 k 3 6.6 pg = 3 ug genomic DNA), each control (5% of the reaction, eluted into 6 mL) should yield a concentration of 30 ng/mL.
For flow-sorting, single cell or nuclei should be easily distinguished based on DAPI signal (on a linear scale) from debris (no signal) and from clumps of multiple cells or nuclei (double, triple, or even higher signal; should be a minority of events) ( Figure 2). We additionally use a typical ''singlet'' gate based on FSC signal height (or width on some sorters).
For whole-genome amplification (WGA), the Tn5 concentration ($0.015 mL per cell) should be titrated so that the PCR product (before size selection) has a relatively flat-slightly higher on the shorter (left) side-length distribution on a Bioanalyzer High Sensitivity DNA chip (i.e., an average length of 500 bp) (Figures 3 and 5A). We typically obtain a concentration of 4 ng/mL both before (eluted into 400 mL per 96-well plate) and after (eluted into 50 mL from half a reaction of a 96-well plate) size selection ( Figure 5).

QUANTIFICATION AND STATISTICAL ANALYSIS
For general diagnosis of whole-genome amplification (WGA)-e.g., mapping rate, genome coverage, and library complexity, please refer to relevant literature (Chen et al., 2017;Huang et al., 2015) for guidelines. Note that CR can be calculated either from numbers before deduplication-i.e., C raw and R raw , or from numbers after deduplication-i.e., C dedup and R dedup . The 2 methods are roughly equivalent because the bias of WGA and sequencing and the distribution of contacts are relatively uniform across the genome and independent of each other. For convenience, we typically use the former.

Success of
A successful reaction should lead to CR R 3% (ideally, R 5%). More aggressive size selection (e.g., 0.6 X rather than 0.7 X) will yield a higher CR (therefore costing less per contact), but at the expense of a lower C dedup (because contacts that are located in shorter fragments are lost during size selection).
CR depends on the length distribution of the final sequencing library, because longer reads are more likely to contain contacts. A metric that depends only on the chromatin conformation capture (3C/Hi-C) step (i.e., independent of WGA) is the ''contact density'' (the average number of contacts per base pair (bp) of genome) CD = CR / L, where L = the average insert size of the sequenced library (in base pairs (bp)).
In particular, because CR is the number of contacts per read pair (i.e., the fraction of contact-containing read pairs among all read pairs) and L is the number of bp per read pair, the formula CD = CR / L gives the number of contacts per bp-i.e., the contact density. Note that L should be calculated from alignments of read pairs that do not harbor contacts-i.e., proper read pairs-rather than from Bioanalyzer traces, because sequencers preferentially read shorter fragments.
The degree of sequencing saturation can be assessed by calculating the ''duplication rate'' (the percentage of contacts that are duplicates) DR = (1 -C dedup / C raw ) 3 100%.
A cell is sequenced to saturation (i.e., sequencing deeper will not lead to many more contacts) if DR R 70% (i.e., on average, each contact is sequenced 3 times). With Nextera amplification, saturation typically occurs with 3-6 m read pairs per cell (i.e., 1 or 2 HiSeq lanes per 96-well plate). At saturation, we typically obtain C dedup = 400 k contacts per cell with Nextera, and C dedup = 1 m with META.
Note that sequencing to saturation is not necessary for certain analysis. For example, clustering of cell types based on single-cell chromatin A/B compartment values (scA/B) may work with as few as 20-50 k contacts per cell, depending on the cell types Tan et al., 2019).
When visualized in Juicebox.js (Robinson et al., 2018), single-cell contact maps should exhibit strong diagonal blocks for intra-chromosomal contacts (same as in bulk Hi-C), and ''patchy'' off-diagonal blobs for inter-chromosomal contacts that are different among cells (a phenomenon unique to single cells). Figure 6A shows contact maps and 3D reconstructions from representative mouse brain cells in comparison with bulk Hi-C; Figure 6B shows a representative t-SNE plot of scA/B (the first 20 principal components) from $2,000 mouse brain cells.

LIMITATIONS
The majority of Dip-C data analysis-including the generation of single-cell chromatin contact maps, calculation of the scA/B matrix, and clustering and identification of cell types-can be performed on any samples. However, reconstruction of 3D structures is limited to normal diploid (requiring a phased SNP file) or haploid cells. In particular, genomic regions with more than 2 copies (i.e., copy number gain) or with 2 identical copies (i.e., loss of heterozygosity (LOH)) cannot be 3D reconstructed. Figure 6. Representative data from the mouse brain (A) Chromatin contact maps (top) and 3D genome structures (bottom) of 2 representative single cells, an aggregation of 795 single cells, and bulk Hi-C. All samples were adult neurons from the mouse brain . Unlike bulk Hi-C, single-cell contact maps show a characteristic pattern of random ''patchiness''-especially for inter-chromosomal contacts-indicating highly heterogenous chromosome interactions among single cells (e.g., each chromosome territory only borders a few others in each cell). Raw bulk Hi-C data was downloaded from (Jiang et al., 2017) and reanalyzed by . Contact maps were visualized with Juicebox.js (Robinson et al., 2018). Note that aggregated or bullk data cannot be represented by a single 3D genome structure, because such data contain mutually conflicting contacts (e.g., inter-chromosomal contacts between all pairs of chromosomes) that is physically impossible for a single structure. (B) t-SNE plot of scA/B showing clusters of 3D genome structure types, from the mouse brain.

TROUBLESHOOTING Problem 1
No amplification product (i.e., nearly 0 ng/mL DNA concentration at step 71) after PCR purification.

Potential solution
If the Positive Control (pure genomic DNA) also failed to amplify, check reagents that are crucial for whole-genome amplification. For example, sufficient MgCl 2 must be present in the Transposition Buffer (step 58) and in the PCR Mix (step 65).
If the Positive Control amplified but cells or nuclei did not, check reagents that are crucial for lysis. For example, Qiagen Protease must be present in the Dip-C Lysis Buffer (step 52). Alternatively, check the flow-sorting procedure (e.g., alignment of the plate) to ensure cells or nuclei are sorted into each well (step 55). Up to 100 cells or nuclei can be sorted or pipetted into a control well for diagnosis.
Problem 2 DNA length is too short or too long (at step 71) after PCR purification.

Potential solution
If DNA length is too short, decrease the concentration of Tn5 transposome (step 59). If decreasing Tn5 does not work, decrease the concentration of Qiagen Protease in the Dip-C Lysis Buffer (step 52, and/or the Stop Mix) to combat lot variation in Qiagen Protease. If DNA is too long, perform the opposite adjustment.
For first-time users, the Tn5 concentration can be roughly titrated with the Positive Control (pure genomic DNA) for maximum DNA yield, and finely titrated with cells or nuclei to obtain the desired DNA length (i.e., 500 bp on average) (Figure 3).
If flow-sorting single cells or nuclei is challenging for titration, a large number of cells or nuclei can be lysed together in the Dip-C Lysis Buffer. The aliquoted lysate (stored indefinitely at À80 C) can be used in place of single cells or nuclei for convenient titration ( Figure 3B).

Problem 3
After sequencing (step 74), some of the 96 wells on a plate yield very few read pairs.

Potential solution
Check the flow-sorting procedure (e.g., alignment of the plate) to ensure cells or nuclei are sorted into each well (step 55). If the plate was not aligned, the flow sorter may miss entire rows or columns of wells.

Potential solution
A low CR indicates that despite successful whole-genome amplification, the chromatin conformation capture step did not generate sufficient contacts. Typically, CR and CD are relatively uniform among cells or nuclei from the same 3C/Hi-C reaction, and vary mostly between reactions and/or cell types.
Check the Digestion Control and Ligation Control (step 47)-the problem is typically caused by insufficient digestion. Switch to another restriction enzyme (e.g., NlaIII usually leads to better digestion), a combination of restriction enzymes, or another 3C/Hi-C protocol or kit. Alternatively, redoing the experiment may lead to a better batch. It's common in 3C/Hi-C to sequence multiple replicates shallowly (step 74), and proceed with those with higher CR.
Note that certain cell types may be especially challenging. For example, rod photoreceptors (Tan et al., 2019) yield fewer contacts per cell, whereas sperm yield almost no contacts per cell.

Potential solution
If possible, obtain more contacts per cell. For example, if the library has not been sequenced to saturation (quantification and statistical analysis; duplication rate DR < 70%), sequence deeper. Alternatively, switch to another restriction enzyme (e.g., NlaIII usually leads to better digestion), a combination of restriction enzymes, or another 3C/Hi-C protocol or kit to increase the contact rate CR. Switch to META from Nextera to detect more contacts.
Alternatively, reduce the resolution of 3D modeling (e.g., from 20 kb to 100 kb). Fewer contacts are required to determine a lower-resolution structure.
Note that certain cell types and chromosome configurations may be especially challenging for existing 3D-modeling algorithms. 3D modeling (e.g., the choice of energy function and energy-minimization procedure) is an area of active research. For example, cells with more inter-chromosomal contacts tend to be easier to model (Stevens et al., 2017). Cells with complex nuclear shapes (e.g., rings, multiple lobes) may yield high RMSD. For diploid cells, low SNP density (e.g., female DBA/2J mice has very few heterozygous SNPs on Chr X) and homolog interactions (e.g., the 2 copies of human Chr 19 both prefer the nuclear center, and may thus interact by chance) may lead to poor 3D modeling of certain regions.

RESOURCE AVAILABILITY
Lead contact Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Longzhi Tan (tttt@stanford.edu).

Materials availability
This study did not generate new unique reagents.

Data and code availability
The code generated during this study is available at GitHub (https://github.com/tanlongzhi/dip-c).
ACKNOWLEDGMENTS I thank D. Xing, C.H. Chang, and X.S. Xie for co-inventing Dip-C and H. Li for co-developing its software packages. I also thank my current advisor, K. Deisseroth, and the three reviewers. L.T. was supported by a School of Medicine Dean's Postdoctoral Fellowship and a Walter V. and Idun Berry Postdoctoral Fellowship from Stanford University.

AUTHOR CONTRIBUTIONS
L.T. wrote the manuscript.

DECLARATION OF INTERESTS
L.T. is an inventor on a patent application (US16/615,872) filed by Harvard that covers Dip-C.

ll
OPEN ACCESS